Where circuit density and performance do not have to be fully maximized in an integrated circuit, it is sometimes found expedient to include in the circuit design additional connections, within the circuit wiring, that can be severed, as desired, at a later stage of the manufacturing process. In this way the circuit can be either repaired by removing bad components (assuming redundancy has been built into the system) or, more generally, a set of general purpose circuits can be personalized to perform any of a range of specific functions.
With the advent of the copper damascene process, it has been a natural step to make the fuses out of copper since making the fuse could be combined with making normal wiring. An example of such a prior art structure is schematically illustrated in FIG. 1. A layer of field oxide 17 is seen to be covering a silicon substrate including, primarily, a P-well 18. The two fragments 16a and 16b of polysilicon wiring would normally be connected to two electrically separate devices but in the fuse arrangement, each is upwardly connected to three levels of dual damascene structures 11 that are embedded in dielectric 10. These share a common trench (or stripe) portion 12 at their topmost level. This shared topmost level of damascene wiring can thus be used as a fuse.
In order to make connecting copper stripe 12 fusible under the influence of externally directed radiation, most of dielectric material 10 directly above 11 is etched away to form fuse cavity 12. A small thickness (typically about 8,000 Angstroms) 14 of dielectric is left in place. This residual layer is thick enough to passivate the copper while at the same time being thin enough to allow the fusing beam to pass through with minimum attenuation as well as being thin enough to disintegrate when the fuse is blown i.e. when 12 becomes discontinuous.
There are several problems associated with this method of making a fusible connection within a copper damascene technology. Because copper is a good thermal conductor and because it has a relatively high melting point, the energy needed to fuse the copper is relatively high--typically about 10 microjoules. This makes copper fuses incompatible with existing laser fusing technologies that were developed prior to the wholesale introduction of the copper damascene technology.
An additional problem associated with copper fuses has been the poor passivation qualities of copper oxide. The latter forms on the bottom surface of window 15 after the fuse has been blown and provides a ready source of copper that can rapidly diffuse through the dielectric down to the silicon devices level where it `poisons` them. Of course, the original damascene structure of which the fuse was a part included barrier layers (not shown) that isolated the copper from all other materials, but these protective layers are no longer present in the vicinity of a blown fuse.
Also seen in FIG. 1 is guard ring 13. This is series of dummy dual damascene structures that form a ring completely surrounding fuse cavity 15. Its purpose is to terminate the propagation of cracks in dielectric 10 that may appear after a fuse has been blown. While microcracks will propagate relatively easily through the hard and brittle dielectric, they cease to grow once they encounter the soft metal of the guard ring. We have shown the guard ring for the sake of completeness and, although it is advisable to use one in conjunction with a fuse structure, it has no direct bearing on the present invention.
A routine search of the prior art was conducted but no references teaching the exact process or structure of the present invention were found. Several prior art references that were encountered were, however, found to be of interest. For example, in U.S. Pat. No. 5,801,094, Yew et al. show a copper dual damascene method.
Mitwalsky et al. In U.S. Pat. No. 5,776,826 show a fuse in an upper level metal layer and introduce an extra level of metal below the fuse to prevent propagation of cracks that might form when a fuse is blown. This essentially the protection ring mentioned above.
In U.S. Pat. No. 5,795,819 Motsiffet al. describe a process in which the same material is used for both a C4 barrier layer and a fuse. Aluminum is mentioned as one of the materials which can be used in this way. It is, however, important to note that there is no other material present between the aluminum and the copper. As will be seen, the present invention teaches the importance of inserting another metal, such as chromium, between the aluminum and the copper.
In U.S. Pat. No. 5,589,706 Mitwalsky et al. show a fuse link and dummy structures for dual damascene process and BEOL (back end of line). Lee et al. (in U.S. Pat. No. 5,757,060,) and Yaung et al. (in U.S. Pat. No. 5,926,697) show a guard rings around a fuse while in U.S. Pat. No. 5,729,041, Yoo et al. show a protective film for a fuse window.